Skip to content

Diagnosis and classification

TL;DR — AN is diagnosed from restrictive intake/low weight, weight-gain fear or weight-preventing behaviour, and disturbed weight/shape experience or recognition of seriousness; no laboratory test confirms it. Atypical AN meets the psychological/behavioural pattern without the low-weight criterion and can show comparable psychopathology and clinically important instability (Walsh 2023, PMID 36508318). BMI is contextual evidence, not a stand-alone diagnostic or admission rule. Bulimia nervosa, binge-eating disorder and ARFID are different diagnoses and appear here only as differential diagnoses.

Core formulation

Question Evidence sought Common error
Is intake persistently restricted? Dietary pattern, avoidance, rituals, trajectory Equating restriction with absent appetite
Is weight significantly low for context? Adult BMI plus history; child/adolescent growth trajectory Applying one universal threshold
Is gain feared or prevented? Stated fear, exercise, concealment, rituals, purging Requiring explicit verbal fear
Is self-evaluation disturbed? Weight/shape overvaluation, body experience, recognition of risk Assuming insight is all-or-none
Is another process primary? Medical, psychiatric and other eating-disorder differential Treating diagnosis as exclusion-free

DSM-5 removed amenorrhoea as a requirement and reduced reliance on a single numeric threshold; contemporary reviews emphasize developmental and longitudinal assessment (Moskowitz 2017, PMID 28532965; Herpertz-Dahlmann 2015, PMID 25455581). ICD and DSM wording differs, so research cohorts should report the exact system and version.

Atypical anorexia nervosa

Atypical AN is classified under other specified feeding or eating disorder when all AN criteria are met except that weight remains within or above a conventionally “normal” range after substantial loss. In a 24-publication systematic review, eating-disorder psychopathology was as high or higher than in AN, non-eating-disorder psychopathology was similar, and many physiological complications occurred, although some were less frequent; evidence on course, treatment response and fully operational criteria remained sparse (Walsh 2023, PMID 36508318). An invited update by the same group, covering 64 publications and published in 2026, confirms and extends this: individuals with atypical AN differ demographically from those with AN and show greater eating-disorder psychopathology, with menstrual disturbance and reduced bone mineral density present but less frequent. Its closing position is that these findings "continue to support the similarities between atypical AN and AN" while underscoring the need for specific diagnostic criteria and for longitudinal studies of course and outcome (Lee 2026, PMID 42557659). Three years and forty extra publications have therefore strengthened the cross-sectional case and left the longitudinal gap untouched.

Dimension Typical AN Atypical AN Interpretation
Current low-weight criterion Present Absent Does not measure rate or magnitude of loss
Restriction/fear/overvaluation Present Present Core psychopathology may be comparable
Bradycardia/orthostasis/electrolytes Possible Possible Assess directly; do not infer from body size
Evidence base Larger 24 comparative publications by 2022, 64 by 2026 — all cross-sectional Treatment extrapolation is common but incompletely tested (Walsh 2023, PMID 36508318; Lee 2026, PMID 42557659)
Bias risk Severity recognized at low weight Delayed recognition in higher-weight bodies Weight stigma can shape referral and care

Two lived-experience analyses of this classification have been published in the same journal, one making the explicit case for re-conceptualizing the “atypical” category (Verma 2024, PMID 37897094) and one offering a lived-experience perspective on how atypical AN is classified (Harrop 2023, PMID 36577133). Neither carries an abstract in PubMed and neither full text was accessible in the 2026-09-02 audit session, so their arguments are cited here only at the level their titles support. The critique they represent — that classification language has consequences for how seriously restriction is taken — is evidence about the effects of a label, not proof that every biological risk is identical.

Subtypes and severity

The restricting and binge-eating/purging subtypes describe recent behaviour; movement between them limits their use as fixed etiological categories. DSM-5 adult severity specifiers use BMI bands, but treatment-outcome work has questioned whether these categories carry sufficient prognostic information. In 128 adult women with AN (64 outpatients, 64 inpatients) treated with CBT-E and sub-categorized by the four DSM-5 severity levels, there were no significant differences across severity groups in either "weight recovery" (BMI ≥18.5 kg/m²) or "good outcome" (BMI ≥18.5 plus minimal residual eating-disorder psychopathology), at end of treatment or at 6- and 12-month follow-up (Dalle Grave 2018, PMID 30059831). Report continuous BMI, percentage of median BMI when developmentally appropriate, rate and magnitude of loss, and physiological findings rather than severity label alone.

Case finding and screening

No laboratory test confirms AN, so case finding depends on questionnaires and clinical suspicion. The five-item SCOFF is the most studied instrument. A meta-analysis of 25 validation studies estimated pooled sensitivity 0.86 (95% CI 0.78–0.91) and specificity 0.83 (95% CI 0.77–0.88), but with a structural caveat: sensitivity was highest in case-control studies of young women with AN and bulimia nervosa, and lower in studies including more men, including binge-eating disorder, or recruiting from large community samples. No included study used a reference standard covering all DSM-5 eating disorders, and the authors concluded there was insufficient evidence to use the SCOFF for the full DSM-5 range in primary care and community settings (Kutz 2020, PMID 31705473).

The independent USPSTF evidence review reached compatible numbers on adult accuracy — SCOFF at a cut point ≥2 had pooled sensitivity 84% (95% CI 74–90%) and specificity 80% (95% CI 65–89%) across 10 studies (n=3,684) — and a harsher conclusion about what that accuracy buys. Across 57 included studies (N=10,773), no study directly evaluated the benefits and harms of screening, and none of 40 intervention RCTs enrolled a screen-detected population (Feltner 2022, PMID 35289875). The Task Force therefore issued an I statement: the evidence is insufficient to assess the balance of benefits and harms of screening for eating disorders in adolescents and adults with normal or high BMI, explicitly excluding people who are underweight or already showing physical signs (Davidson 2022, PMID 35289876).

The gap this exposes is specific and worth stating plainly: an instrument with 84% sensitivity is being used to find a population for which no trial has ever tested whether finding them earlier by screening improves outcome. That is different from saying screening is useless, and different from the separate, better-supported case for reducing delay in people who have already presented — see service models and setting.

ICD-11 versus ICD-10, and the residual-category problem

The DSM-5 changes described above have a parallel in ICD-11, and unusually for a classification revision, the change was tested empirically before publication. In a vignette-based, randomized field study of 2,288 clinicians from WHO's Global Clinical Practice Network conducted in Chinese, English, French, Japanese and Spanish, the proposed ICD-11 guidelines significantly improved diagnostic accuracy over ICD-10 for every feeding-and-eating disorder tested and attained higher clinical-utility ratings, with similar results across all five languages; adding binge-eating disorder and ARFID as named categories reduced reliance on residual diagnoses (Claudino 2019, PMID 31084617).

Applying both systems to the same patients shows what that means at the clinic level. In 82 patients aged 0–17 assessed with the Eating Disorder Examination, its child version and the EDE ARFID module, the number of residual restrictive eating disorders fell significantly from ICD-10 to ICD-11 because patients crossed over into full-threshold diagnoses — principally AN or ARFID. Those reclassified to ARFID were younger, had earlier onset, more restrictive eating and more somatic comorbidity than those reclassified to AN; patients who remained residual under both systems were younger, had earlier onset, less shape concern and more somatic comorbidity (Düplois 2023, PMID 37243388). Classification change therefore moves people between categories in a patterned rather than a random way, which matters for any comparison of prevalence or outcome across diagnostic eras.

Do the categories hold over time? Two answers

This is one of the clearest unresolved disagreements in AN nosology, and the two sides use different designs.

Study Design Finding
Eddy 2008, PMID 18198267 216 women with AN or BN followed 7 years with weekly symptom data (Eating Disorder Longitudinal Interval Follow-Up Examination) The majority of women with AN crossed diagnoses: more than half moved between restricting and binge-eating/purging subtypes, and one third crossed to bulimia nervosa (though likely to relapse back to AN). Women with BN rarely crossed to AN. Conclusion: the longitudinal AN/BN distinction is supported; the AN subtyping scheme is not
Schaumberg 2019, PMID 29911514 9,622 individuals in Swedish eating-disorder quality registers seen at least twice, 1999–2013, two birth cohorts Diagnostic instability was common, but transition between threshold diagnoses was infrequent; transitions to remission typically followed a diagnostic state matching the initial diagnosis or a subthreshold state. Younger cohort more likely to reach remission. Conclusion: "more temporal continuity in eating disorder presentations than suggested by previous research"

The two are not straightforwardly reconcilable. Weekly prospective interview data in a specialist longitudinal cohort find crossover to be the norm; register data on a population-scale treatment-seeking sample find threshold-to-threshold transition uncommon. Measurement frequency, diagnostic granularity and ascertainment differ enough that both could be locally correct — a register records what was coded at a contact, an interview records what was true each week. What follows for this wiki is that subtype should be reported as a state at a stated time, and studies comparing "restricting" and "binge-purge" cohorts should state how and when subtype was determined.

Differential diagnosis

Candidate Distinguishing question Boundary in this wiki
Bulimia nervosa Are recurrent binge/purge episodes primary without persistently low weight? Differential only
Binge-eating disorder Are recurrent binges present without regular compensation/restriction syndrome? Differential only
ARFID Is restriction driven by sensory features, low interest or feared consequences rather than weight/shape concerns? Differential only
Major depression Is reduced intake secondary to anhedonia/appetite loss without AN cognitions? Assess both; comorbidity is possible
OCD Are rituals broader and not principally organized around weight/shape/intake? Shared traits do not collapse diagnoses
Gastrointestinal/endocrine disease Is weight loss driven by malabsorption, inflammation, endocrine excess/deficiency or other disease? Medical evaluation remains necessary
Relative energy deficiency in sport Is low energy availability present; are AN cognitions/behaviours also present? Syndromes can overlap

Medical mimics can include thyroid disease, adrenal disease, inflammatory bowel disease, coeliac disease, chronic infection, malignancy and diabetes; a focused history, examination and testing follow the clinical presentation (Harrington 2015, PMID 25591200). Diagnosis of AN does not immunize against coexisting organic disease. The Society for Adolescent Health and Medicine frames the assessment around restrictive eating disorders rather than around a diagnostic label, explicitly including adolescents and young adults who are not underweight — a framing that makes atypical AN a routine part of medical evaluation rather than an exception to it (PMID 36058805).

Instrument thresholds are population-dependent

The Eating Disorder Examination-Questionnaire supplies the "eating-disorder psychopathology within population range" component of most composite recovery definitions, and its norms are not weight-neutral. In 3,000 women aged 16–50 randomly drawn from the Norwegian National Population Register, mean global EDE-Q was 1.27 (SD 1.19); scores fell with age and rose with BMI, and in regression BMI accounted for 19% of global EDE-Q variance against 2% for age. Among women with overweight or obesity, 30–40% scored above the recommended clinical cut-off for Shape and Weight Concern (Rø 2012, PMID 22365803). Applying a single EDE-Q cut-off across the weight range therefore misclassifies in a patterned direction — and does so precisely in the higher-weight population where atypical AN is under-recognized.

The ARFID boundary is not clean

DSM-5 separates ARFID from AN by the absence of weight/shape overvaluation driving restriction. Latent profile analysis of 202 treatment-seeking individuals aged 10–79 with ARFID or a non-ARFID eating disorder, using the Nine-Item ARFID Screen (Picky, Appetite, Fear subscales) and the EDE-Q Restraint subscale, produced five profiles. Three behaved as DSM-5 predicts: two endorsing solely ARFID motivations (82% and 68% with ARFID diagnoses) and one endorsing solely restraint (11% ARFID). But a Restraint/ARFID-Mixed profile (n=24) endorsed both motivation types and was 92% non-ARFID eating-disorder diagnoses, comprising 18% of all non-ARFID cases in the sample (Abber 2024, PMID 38801097). Roughly one in five people with a non-ARFID eating disorder carries substantial ARFID-type restriction motivations, which the current diagnostic boundary treats as mutually exclusive.

Assessment dataset

Domain Minimum research/clinical record
Anthropometry Height, weight, BMI, historical maximum/minimum, rate/magnitude of change, growth curve
Intake Pattern, exclusions, fluids, supplements, food insecurity
Behaviours Exercise, vomiting, laxatives, diuretics, water loading, weighing/checking
Physiology Pulse, blood pressure including orthostasis, temperature, hydration, ECG and indicated laboratories
Mental state Weight/shape beliefs, motivation, depression, anxiety, OCD symptoms, self-harm/suicide risk
Context Development, family/carer setting, trauma, culture, gender identity, sport, access barriers

Open questions

  • What operational definition of “significant weight loss” best predicts instability and outcome in atypical AN (Walsh 2023, PMID 36508318)?
  • Do current BMI severity specifiers add prognostic information beyond trajectory and physiology (Dalle Grave 2018, PMID 30059831)?
  • Which classification language reduces delayed care without erasing clinically relevant differences (Harrop 2023, PMID 36577133)?
  • Does screening for eating disorders improve outcomes? No study has directly evaluated the benefits and harms of screening, and no treatment trial has enrolled a screen-detected population (Feltner 2022, PMID 35289875; Davidson 2022, PMID 35289876).
  • Is diagnostic crossover the rule or the exception? Prospective weekly follow-up says the majority of AN cases cross diagnoses over 7 years; a 9,622-person register study says threshold-to-threshold transition is infrequent (Eddy 2008, PMID 18198267; Schaumberg 2019, PMID 29911514).
  • How much of any observed change in AN prevalence across diagnostic eras is reclassification rather than incidence, given that ICD-11 moves patients out of residual categories into AN and ARFID in a patterned way (Claudino 2019, PMID 31084617; Düplois 2023, PMID 37243388)?
  • Should EDE-Q cut-offs used in recovery definitions be BMI-stratified, given that BMI explains 19% of variance in global score in the general population (Rø 2012, PMID 22365803)?
  • Are ARFID-type and shape/weight-type restriction motivations mutually exclusive, as DSM-5 assumes, when 18% of non-ARFID eating-disorder cases endorse both (Abber 2024, PMID 38801097)?

References

  1. Walsh BT, et al. A systematic review comparing atypical anorexia nervosa and anorexia nervosa. Int J Eat Disord. 2023. PMID 36508318.
  2. Moskowitz L, Weiselberg E. Anorexia nervosa/atypical anorexia nervosa. Curr Probl Pediatr Adolesc Health Care. 2017. PMID 28532965.
  3. Herpertz-Dahlmann B. Adolescent eating disorders: update on definitions, symptomatology, epidemiology, and comorbidity. Child Adolesc Psychiatr Clin N Am. 2015. PMID 25455581.
  4. Dalle Grave R, et al. DSM-5 severity specifiers for anorexia nervosa and treatment outcomes in adult females. Eat Behav. 2018. PMID 30059831.
  5. Harrop EN, et al. A lived experience perspective on the classification of atypical anorexia nervosa. Int J Eat Disord. 2023. PMID 36577133.
  6. Verma S, et al. A case for re-conceptualizing the “atypical”—a lived experience perspective. Int J Eat Disord. 2024. PMID 37897094.
  7. Harrington BC, et al. Initial evaluation, diagnosis, and treatment of anorexia nervosa and bulimia nervosa. Am Fam Physician. 2015. PMID 25591200.
  8. Society for Adolescent Health and Medicine. Medical management of restrictive eating disorders in adolescents and young adults. J Adolesc Health. 2022. PMID 36058805.
  9. Lee V, Hagan KE. An invited updated systematic review and meta-analysis comparing atypical anorexia nervosa and anorexia nervosa. Int J Eat Disord. 2026. PMID 42557659.
  10. Kutz AM, et al. Eating disorder screening: a systematic review and meta-analysis of diagnostic test characteristics of the SCOFF. J Gen Intern Med. 2020;35:885-893. PMID 31705473.
  11. Feltner C, et al. Screening for eating disorders in adolescents and adults: evidence report and systematic review for the US Preventive Services Task Force. JAMA. 2022;327:1068-1082. PMID 35289875.
  12. US Preventive Services Task Force (Davidson KW, et al.). Screening for eating disorders in adolescents and adults: USPSTF recommendation statement. JAMA. 2022;327:1061-1067. PMID 35289876.
  13. Claudino AM, et al. The classification of feeding and eating disorders in the ICD-11: results of a field study comparing proposed ICD-11 guidelines with existing ICD-10 guidelines. BMC Med. 2019;17:93. PMID 31084617.
  14. Düplois D, et al. Distribution and clinical comparison of restrictive feeding and eating disorders using ICD-10 and ICD-11 criteria. Int J Eat Disord. 2023;56:1717-1729. PMID 37243388.
  15. Eddy KT, et al. Diagnostic crossover in anorexia nervosa and bulimia nervosa: implications for DSM-V. Am J Psychiatry. 2008;165:245-250. PMID 18198267.
  16. Schaumberg K, et al. Patterns of diagnostic transition in eating disorders: a longitudinal population study in Sweden. Psychol Med. 2019;49:819-827. PMID 29911514.
  17. Rø Ø, et al. The impact of age and BMI on Eating Disorder Examination Questionnaire (EDE-Q) scores in a community sample. Eat Behav. 2012;13:158-161. PMID 22365803.
  18. Abber SR, et al. Latent profile analysis reveals overlapping ARFID and shape/weight motivations for restriction in eating disorders. Psychol Med. 2024;54:2956-2966. PMID 38801097.